Remote Site Design Question

[sosborne]

Ultimately the plan is to have two physical Win 2016 Servers (one at primary site and one at DR site) with direct attached storage (160TB). Currently we have 1 at the primary and I am building up the server for the DR site. The idea is that B&R server at each site will back up the VMs on the local site, and that a backup copy job will copy the short term and the long term backups from primary to DR and vice versa (for the few VMs there) after an initial seeding. And once that is all setup I will look to replicate key VM's by setting up what I have heard described as pull replication at the remote site (and possibly the other way if needed) especially for the single vCentre we have at the primary site.

Now my big question is what is the best way to go about the install at the remote site? From reading documentation and forum posts I get the feeling there are two potential ways to do this either as a separate instance at the remote site or as a part of the backup infrastructure already setup at the primary. My gut says to go with a separate instance especially as we want to be able to disaster recovery, replication and seamless failover/roll back. Is this the best way to go? And who has similar setups, how did you configure it and what has been your experience with it? Furthermore how does B&R Enterprise manager fit into the picture?

[foggy]

Hi Simon, your understanding is correct. If you're willing to have pull replication and be able to failover via Veeam B&R console in case of DR, then the second remote instance is the way to go. Enterprise Manager will provide you a single pane of glass across both instances.

[sosborne]

Thanks for the response.

Is managing the fail-over process via the B&R console the preferred trade off as opposed to avoiding site specific dependencies if the primary site were to fail? Is it possible to automate fail-over to any degree if there are two instances?

Alternatively, is it possible to set up the backup infrastructure in such a way that a single instance with extra components in certain places (e.g extra proxies, repositories etc) would avoid the dependencies on a primary site but give you the advantages of a single install?

[foggy]

You can certainly make do with a single Veeam B&R server instance located in secondary site, just make sure you have your proxy servers close to the source storage for faster data retrieval. You can also have an additional Veeam B&R console installed in the primary site for more convenient access to the backup server.

[sosborne]

Thanks for fleshing out the single instance option. I guess I would also need a repository server at the primary site for the local storage there?

What option is preferred to achieve the goal of backup and replication at each site whilst avoiding dependencies and dealing with an awkward replication/fail-over process? I know that no option is ideal (to all factors) and that compromises will be made but as long as I am making an informed decision the that's all I can ask.

[foggy]

Sure, local repository is always preferred.

In case of cross-site replication, the only way to fully avoid dependencies is to have backup server in some third location (otherwise, if the site where backup server is located, is lost, you will not be able to failover quickly via Veeam B&R). But it is completely up to you to decide what configuration is applicable and reasonable in your case. For example, if VMs running in DR site are not that critical and you can allow longer RTO for them, there's no need in fast automatic failover - just guessing... firing up a new Veeam B&R instance and importing configuration doesn't usually take much time).

[sosborne]

It has been useful to discuss the intricacies of the solution as it shows how particular they are to your infrastructure. Between this conversation with yourself and another Veeam engineer it has made me realise how flexible you can be with the infrastructure to meet very particular setups.

The idea put forward is to have a Virtual VBR server hosted at primary and replicated to DR (so that should the site to site link go down we can still backup, and likewise still function should the primary site go down) with proxy and repository roles installed on each physical server. The idea seems to me to fit the purpose, so just have to think how to make it happen. I guess install proxy and repository roles on 2nd server first, create a Virtual VBR, and then figure out how to export/import the config/DB to the virtual system.

[sosborne]

Out of interest does anyone know if the Backup Server needs to be Win 2016 if the servers acting as proxy/repositories that are actually attached to the disks are already Win 2016?

[DGrinev]

Hi Simon,

Nope, you can use a mix of supported Windows versions for different backup components. Thanks!

[sosborne]

Thanks. The reason I asked is that currently the backup server is Win 2016 with ReFS. Is it possible to have a Virtual Win 2012 R2 backup server and Win 2016 proxy/repository with ReFS?

[DGrinev]

Sure, feel free to stick with the plan.
Also, you can review a few existing discussions on the forum like this one - "...Server 2012 R2 or 2016". Thanks!